An anticipated update has dropped into my M365 tenant today for Microsoft Syntex image tagging. Once SharePoint libraries are enabled for Syntex Image tagging and images are uploaded they are then auto scanned by AI. The images are then tagged with descriptive keywords with no training or manual tagging required.
The keywords are stored in a special Image Tags managed metadata column – which means once the images are tagged you can then search, refine and identify images.
This is currently only available if Syntex pay as you go is configured in your tenant & linked to an Azure subscription. Currently billing is charged per image processed whereby each image counts as one transaction and it is currently priced at $0.001 per image processed. Syntex image tagging is only charged in libraries where Syntex image tagging is explicitly enabled and only on new content added to the libraries.
I believe it may use Azure AI Vision behind the scenes but Syntex wraps it all up in a easy way to have images auto tagged in a library without any code/training involved!
Setup Syntex Image Tagging
Below I will walk you through the process of enabling Syntex Image tagging in your tenant and testing it with adding images.
Firstly ensure Syntex Pay As You Go is setup and linked to an Azure Subscription this can be done in the M35 Admin Centre with the Syntex Admin Centre (Link).
In the M365 Syntex admin centre ensure your Azure subscription is setup and then click on Manage Microsoft Syntex.
Click on Image Tagging
Here are the current configuration options to either allow Syntex image tagging to be used in all libraries in all SharePoint sites or no libraries at all.
I will enable Syntex Image Tagging on all libraries in all SharePoint sites.
NOTE – there is currently no option to enable Syntex image tagging on just specific sites – it is an all (sites) or nothing (no sites) approach.
I then created a new library in a SharePoint site and called it Image Tagging. To enable Syntex Image tagging I can then then go to the Document Library menu and click on the Automate drop down and select Configure Image Tagger.
I am then prompted to Enable Image tagger on this library.
Once enabled a special Image Tags column is added to the library and I can then go and view the settings of the column. I see there is a special configuration setting enabled on the column to “Automatically tag images with detected objects”. If you already have an Image Tags column in your library you can use these steps to enable Image tagging and use the existing column.
I then added some images to the library and waited for them to be processed. Currently it can be anytime from 5 minutes up to a maximum of 24 hours before the images are tagged.
Image tagging is currently available for all common mage file types: .bmp, .png, .gif, .jpeg, .jpg, .tif etc and includes lots of obscure image files types (see full list of supported file types).
My uploaded images are then tagged with descriptive keywords (see image below). It seemed to take around 30 minutes for the images to be fully processed/tagged by the AI. I will analyse the tags more in the summary but you can if you like modify the existing tags to add more descriptive tags or corrections.
TIP: I also added an Image column to my library and called it Thumbnail – once this is added it automatically presents a preview of the file – so you can see a preview of the image. This would be a great feature to automatically add.
You then have the ability to use search to search for any of the tags identified for the images.
Summary
Syntex Image tagging seems to work well to descriptively label image with keywords. The idea then you can then use the keywords to group images, find images and generally identify them. This will be very powerful and help employees to find images.
The tags are however quite generic at the moment and will struggle to identify content in less common images i.e. who is the famous person in photo, company in the logo, what is the drawing of etc. I can however see Syntex image tagging getting better and better and operate more like other AI models such as GPT-4 that can process and identify content in the images. Along with perhaps providing industry specific image tagging AI – i.e. image tagging catered to construction, medical etc to identify common images in these industries.
Looking forward to seeing this in use with organisations and images appearing better in search. So not just having to use filename.png as a search term or relying on others to manually add metadata to help find an image.
Leon, thank you for the detailed breakdown on using this paid service. Don’t think I see a lot of value in it yet when the “AI” doesn’t even seem to apply simple OCR technology in the case of the Leicester City Football Club emblem. Keep us posted as it matures.